Consistency of Discrete Nonparametric Priors for Continuous Data
Pierpaolo De Blasi, Antonio Lijoi, Igor Pruenster
University of Turin, Italy; University of Pavia, Italy; University of Turin, Italy

Most of the discrete nonparametric priors currently use, with the exception of the Dirichlet processes, are inconsistent if used to model directly continuous data. On the other hand, they generally have large enough weak support to make them suitable for hierarchical mixture modelling of continuous models. In this paper we provide sufficient condition for consistency for Gibbs-type prior and present two examples within this class which exhibit completely opposite behaviour in the limit.

Keywords: Bayesian nonparametrics; Consistency; Gibbs-type priors

Biography: I am assistant professor in the Department of Statistics and Applied Mathematics at the University of Turin, Italy. I graduated with a PhD in Statistics from Bocconi University, Milan, in April 2006 with a thesis in Bayesian semi-parametric methods under the supervision of Nils Hjort of the University of Oslo. After the PhD, I spent one year as post-doctoral research associate in the Department of Biology at the University of Oslo.